diff mbox series

[V2,1/4] blk-mq: add API of blk_mq_unfreeze_queue_force

Message ID 20230620013349.906601-2-ming.lei@redhat.com (mailing list archive)
State New, archived
Headers show
Series nvme: fix two kinds of IO hang from removing NSs | expand

Commit Message

Ming Lei June 20, 2023, 1:33 a.m. UTC
NVMe calls freeze/unfreeze in different contexts, and controller removal
may break in-progress error recovery, then leave queues in frozen state.
So cause IO hang in del_gendisk() because pending writeback IOs are
still waited in bio_queue_enter().

Prepare for fixing this issue by calling the added
blk_mq_unfreeze_queue_force when removing device.

Signed-off-by: Ming Lei <ming.lei@redhat.com>
---
 block/blk-mq.c         | 28 +++++++++++++++++++++++++---
 block/blk.h            |  3 ++-
 block/genhd.c          |  2 +-
 include/linux/blk-mq.h |  1 +
 4 files changed, 29 insertions(+), 5 deletions(-)
diff mbox series

Patch

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 16c524e37123..4c02c28b4835 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -185,26 +185,48 @@  void blk_mq_freeze_queue(struct request_queue *q)
 }
 EXPORT_SYMBOL_GPL(blk_mq_freeze_queue);
 
-void __blk_mq_unfreeze_queue(struct request_queue *q, bool force_atomic)
+void __blk_mq_unfreeze_queue(struct request_queue *q, bool force_atomic,
+		bool force)
 {
 	mutex_lock(&q->mq_freeze_lock);
 	if (force_atomic)
 		q->q_usage_counter.data->force_atomic = true;
-	q->mq_freeze_depth--;
+	if (force) {
+		if (!q->mq_freeze_depth)
+			goto unlock;
+		q->mq_freeze_depth = 0;
+	} else
+		q->mq_freeze_depth--;
 	WARN_ON_ONCE(q->mq_freeze_depth < 0);
 	if (!q->mq_freeze_depth) {
 		percpu_ref_resurrect(&q->q_usage_counter);
 		wake_up_all(&q->mq_freeze_wq);
 	}
+unlock:
 	mutex_unlock(&q->mq_freeze_lock);
 }
 
 void blk_mq_unfreeze_queue(struct request_queue *q)
 {
-	__blk_mq_unfreeze_queue(q, false);
+	__blk_mq_unfreeze_queue(q, false, false);
 }
 EXPORT_SYMBOL_GPL(blk_mq_unfreeze_queue);
 
+/*
+ * Force to unfreeze queue
+ *
+ * Be careful: this API should only be used for avoiding IO hang in
+ * bio_queue_enter() when going to remove disk which needs to drain pending
+ * writeback IO.
+ *
+ * Please don't use it for other cases.
+ */
+void blk_mq_unfreeze_queue_force(struct request_queue *q)
+{
+	__blk_mq_unfreeze_queue(q, false, true);
+}
+EXPORT_SYMBOL_GPL(blk_mq_unfreeze_queue_force);
+
 /*
  * FIXME: replace the scsi_internal_device_*block_nowait() calls in the
  * mpt3sas driver such that this function can be removed.
diff --git a/block/blk.h b/block/blk.h
index 608c5dcc516b..6ecabd844e13 100644
--- a/block/blk.h
+++ b/block/blk.h
@@ -33,7 +33,8 @@  struct blk_flush_queue *blk_alloc_flush_queue(int node, int cmd_size,
 void blk_free_flush_queue(struct blk_flush_queue *q);
 
 void blk_freeze_queue(struct request_queue *q);
-void __blk_mq_unfreeze_queue(struct request_queue *q, bool force_atomic);
+void __blk_mq_unfreeze_queue(struct request_queue *q, bool force_atomic, bool
+		force);
 void blk_queue_start_drain(struct request_queue *q);
 int __bio_queue_enter(struct request_queue *q, struct bio *bio);
 void submit_bio_noacct_nocheck(struct bio *bio);
diff --git a/block/genhd.c b/block/genhd.c
index f71f82991434..184aa968b453 100644
--- a/block/genhd.c
+++ b/block/genhd.c
@@ -708,7 +708,7 @@  void del_gendisk(struct gendisk *disk)
 	 */
 	if (!test_bit(GD_OWNS_QUEUE, &disk->state)) {
 		blk_queue_flag_clear(QUEUE_FLAG_INIT_DONE, q);
-		__blk_mq_unfreeze_queue(q, true);
+		__blk_mq_unfreeze_queue(q, true, false);
 	} else {
 		if (queue_is_mq(q))
 			blk_mq_exit_queue(q);
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index f401067ac03a..fa265e85d753 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -890,6 +890,7 @@  void blk_mq_tagset_busy_iter(struct blk_mq_tag_set *tagset,
 void blk_mq_tagset_wait_completed_request(struct blk_mq_tag_set *tagset);
 void blk_mq_freeze_queue(struct request_queue *q);
 void blk_mq_unfreeze_queue(struct request_queue *q);
+void blk_mq_unfreeze_queue_force(struct request_queue *q);
 void blk_freeze_queue_start(struct request_queue *q);
 void blk_mq_freeze_queue_wait(struct request_queue *q);
 int blk_mq_freeze_queue_wait_timeout(struct request_queue *q,